LONDON, 5th October 2026: Vamstar, a London-based life sciences AI company, today announced that it has secured £900k ($1.2M) funding through the highly selective Innovate UK’s Sovereign AI programme, which selected only the top 1% of projects. The funding will support a UK-owned AI platform that helps health technology developers, anticipate payer requirements and predict launch outcomes before trial protocols are locked in.

The award moves Vamstar's technology from a validated proof of concept to a working demonstrator operating in a real-world environment over the next 12 months. The platform will be offered as a cloud service to pharmaceutical, biotech and medtech companies, NHS bodies and healthcare innovators.

A timing problem, not a science problem
Global pharmaceutical R&D spending is more than £200 billion ($264 billion) a year1, and most candidates that reach late-stage trials never become commercially viable. Many of those failures can be avoided. Evidence strategy and pricing assumptions are fixed at trial design, often years before bodies such as NICE decide whether the evidence justifies the price. By the time a payer finds a gap, the trial can no longer be changed. Patients wait longer, health systems face slower and costlier access decisions, and developers lose value they could have kept.
Vamstar's platform brings payer intelligence into the design stage.
Its agentic AI architecture connects clinical trial evidence, health-economic modelling and real-world market and pricing data in one system. Development teams can see which evidence payers will expect, what price the evidence will support, and which design choices give the best chance of market access, while those choices are still open.
"The industry spends years generating evidence and only learns at the end whether payers will accept it. We are reversing that order" said Praful Mehta, CEO and co-founder of Vamstar. "If developers can see the market-access verdict while the trial can still change, they design better trials, fewer good products fail for the wrong reasons, and patients get them sooner.



Launching a new phase, from proof of concept to demonstrator
The new phase builds on an earlier Innovate UK-funded Sovereign AI proof of concept.

In that project Vamstar showed that HTA evidence retrieval and health-economic modeling could be automated with results consistent with published NICE outcomes with >90% ICER/QALY reproducibility (presented at the ABPI conference in 2026). The demonstrator will:


1. Extend coverage across six therapeutic areas
2. Add clinical evidence, outcomes and real-world market data sources
3. Run at scale as a cloud service in a realistic environment

Built for reliability and transparency

Every output links back to its source evidence. The system runs on UK infrastructure, follows NICE methods and governance requirements, and is being validated independently by several leading academic and commercial partners.
"Within this regulated domain, a solution that cannot provide source attribution, end-to-end traceability and transparent reasoning offers no practical value ," said Dr Richard Freeman, CTO and co-founder of Vamstar. "This is actually the first time that a system interconnects clinical trials, regulatory market data, and the actual post-approval market and commercial worlds together at scale, allowing for a rapid feedback loop. This is powered by a massive agentic AI architecture that combines a large-scale knowledge graph, large language models, and machine learning models."
A similar approach applies to medical devices and diagnostics, where NICE's Early Value Assessment and technology appraisal routes increasingly decide adoption. Vamstar's Polaris platform already supports medtech manufacturers with pricing, tender, and market-access intelligence in more than 100+ countries.
Vamstar is inviting a small number of pharmaceutical, biotech, medtech and NHS organisations to shape early deployment.
Teams preparing HTA submissions or making late-stage development decisions can register interest at [email protected]